Because it is not possible to identify species with echosounders alone, trawling is widely used as a method for collecting species and size composition data for allocating acoustic fish density estimates to species or size groups. In the Laurentian Great Lakes, data from midwater trawls are commonly used for such allocations. However, there are no rules for how much midwater trawling effort is required to adequately describe species and size composition of the pelagic fish communities in these lakes, so the balance between acoustic sampling effort and trawling effort has been unguided. We used midwater trawl data collected between 1986 and 2008 in lakes Michigan and Huron and a variety of analytical techniques to develop guidance for appropriate levels of trawl effort. We used multivariate regression trees and re-sampling techniques to i. identify factors that influence species and size composition of the pelagic fish communities in these lakes, ii. identify stratification schemes for the two lakes, iii. determine if there was a relationship between uncertainty in catch composition and the number of tows made, and iv. predict the number of tows required to reach desired uncertainty targets. We found that depth occupied by fish below the surface was the most influential explanatory variable. Catch composition varied between lakes at depths <38.5 m below the surface, but not at depths ≥38.5 m below the surface. Year, latitude, and bottom depth influenced catch composition in the near-surface waters of Lake Michigan, while only year was important for Lake Huron surface waters. There was an inverse relationship between RSE [relative standard error = 100 × (SE/mean)] and the number of tows made for the proportions of the different size and species groups. We found for the fifth (Lake Huron) and sixth (Lake Michigan) largest lakes in the world, 15–35 tows were adequate to achieve target RSEs (15% and 30%) for ubiquitous species, but rarer species required much higher, and at times, impractical effort levels to reach these targets.